Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.34 as permitted sender) client-ip=23.128.96.34;
From:   isaku.yamahata@intel.com
To:     kvm@vger.kernel.org, linux-kernel@vger.kernel.org
Cc:     isaku.yamahata@intel.com, isaku.yamahata@gmail.com,
        Michael Roth <michael.roth@amd.com>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Sean Christopherson <seanjc@google.com>,
        linux-coco@lists.linux.dev, Chao Peng <chao.p.peng@linux.intel.com>
Subject: [PATCH 03/12] x86/mce/inject: Add notrigger entry to suppress MCE injection
Date:   Tue, 10 Oct 2023 01:35:11 -0700
Message-Id: <97809b68e427922948044e33599c2fc7c9f6134c.1696926843.git.isaku.yamahata@intel.com>
In-Reply-To: <cover.1696926843.git.isaku.yamahata@intel.com>
References: <cover.1696926843.git.isaku.yamahata@intel.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

From: Isaku Yamahata <isaku.yamahata@intel.com>

The current x86 MCE injection framework injects MCE when writing to
/sys/kernel/debug/mce-inject/bank.  KVM wants to inject machine check on
behalf of vcpu context instead of the context of writing to the bank file.

Because ACPI APEI has a similar requirement and it adds
/sys/kernel/debug/apei/notrigger to suppress immediate injection.
By Following it, add /sys/kernel/debug/mce-inject/notrigger to suppress
MCE injection.

The alternative is add new value "notrigger" to
/sys/kernel/debug/mce-inject/flags in addition to "sw", "hw", "df", and
"th".  Because it may break user space ABI, this option follow ACPI APEI
error injection.

Supposed usage flow:
$ echo 1 > notrigger
  ... setup MCE values
$ echo 0 > bank
  The last step to setup mce value to inject with MC bank 0.
  Originally this step injects mce.  With noijnect=1, don't inject.

$ echo 1 > /sys/kernel/debug/kvm/<pid>-<fd>/vcpu<N>/mce-inject
  tell KVM to inject MCE in the context of vcpu.

Signed-off-by: Isaku Yamahata <isaku.yamahata@intel.com>
---
 arch/x86/kernel/cpu/mce/inject.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

diff --git a/arch/x86/kernel/cpu/mce/inject.c b/arch/x86/kernel/cpu/mce/inject.c
index 461858ae18f9..88603a6c0afe 100644
--- a/arch/x86/kernel/cpu/mce/inject.c
+++ b/arch/x86/kernel/cpu/mce/inject.c
@@ -34,6 +34,7 @@
 #include "internal.h"
 
 static bool hw_injection_possible;
+static u64 notrigger;
 
 /*
  * Collect all the MCi_XXX settings
@@ -598,6 +599,8 @@ static int inj_bank_set(void *data, u64 val)
 	}
 
 	m->bank = val;
+	if (notrigger)
+		return 0;
 
 	/*
 	 * sw-only injection allows to write arbitrary values into the MCA
@@ -637,6 +640,21 @@ MCE_INJECT_GET(bank);
 
 DEFINE_SIMPLE_ATTRIBUTE(bank_fops, inj_bank_get, inj_bank_set, "%llu\n");
 
+static int inj_notrigger_get(void *data, u64 *val)
+{
+	*val = notrigger;
+	return 0;
+}
+
+static int inj_notrigger_set(void *data, u64 val)
+{
+	notrigger = val;
+	return 0;
+}
+
+DEFINE_SIMPLE_ATTRIBUTE(notrigger_fops, inj_notrigger_get, inj_notrigger_set,
+			"%llx\n");
+
 static const char readme_msg[] =
 "Description of the files and their usages:\n"
 "\n"
@@ -685,6 +703,9 @@ static const char readme_msg[] =
 "\n"
 "mcgstatus:\t Set MCG_STATUS: the bits in that MSR describes the current state\n"
 "\t of the processor after the MCE.\n"
+"\n"
+"notrigger:\t Suppress triggering the injection when set to non-zero\n"
+"\t The injection is triggered by other way.\n"
 "\n";
 
 static ssize_t
@@ -714,6 +735,8 @@ static struct dfs_node {
 	{ .name = "cpu",	.fops = &extcpu_fops, .perm = S_IRUSR | S_IWUSR },
 	{ .name = "mcgstatus",	.fops = &mcgstatus_fops,
 						      .perm = S_IRUSR | S_IWUSR },
+	{ .name = "notrigger",	.fops = &notrigger_fops,
+						      .perm = S_IRUSR | S_IWUSR },
 	{ .name = "README",	.fops = &readme_fops, .perm = S_IRUSR | S_IRGRP | S_IROTH },
 };
 
-- 
2.25.1